Dies ist eine Übersichtsseite mit Metadaten zu dieser wissenschaftlichen Arbeit. Der vollständige Artikel ist beim Verlag verfügbar.
S752 Artificial Intelligence for Pre-Colonoscopy Patient Guidance: An Evaluation of ChatGPT's Accuracy Against Clinical Guidelines
2
Zitationen
2
Autoren
2023
Jahr
Abstract
Introduction: The digital revolution has transformed healthcare. Patients use the internet and social media for medical advice. Artificial intelligence (AI)-driven chatbots like OpenAI's ChatGPT are becoming increasingly popular sources of medical advice. Given this increasing interest, it is paramount to ensure the reliability and accuracy of information provided by these AI platforms. This study sought to evaluate the ability of ChatGPT to dispense advice aligned with established guidelines in response to patient queries regarding pre-colonoscopy care. Methods: OpenAI's ChatGPT-4 was presented with a series of 25 patient-like questions, which covered four critical aspects of pre-colonoscopy care. These included the recommended diet, the protocol for bowel preparation, the management of cardiovascular medications (emphasizing anticoagulants, antiplatelets, anti-hypertensives, and diabetes medications), and antibiotic prophylaxis in patients with prosthetic heart valves, heart valve infection, and those with implanted devices like pacemakers, implantable cardioverter-defibrillator (ICDs), inferior vena cava (IVC) filters, or prior joint prosthetics. The responses from ChatGPT were compared with guidelines from esteemed organizations, including the European Society of Gastrointestinal Endoscopy (ESGE), United States Multi-Society Task Force (USMSTF), American College of Gastroenterology-Canadian Association of Gastroenterology (ACG-CAG), American College of Cardiology-American Heart Association (ACC-AHA), American Society for Gastrointestinal Endoscopy (ASGE), and the Australian Diabetes Society (ADS). Results: ChatGPT's responses to all 25 inquiries were found to align with the established guidelines. For queries regarding bowel preparation and diet, the responses were consistent with the ESGE and USMSTF guidelines. ChatGPT accurately mirrored the recommendations from the ACG-CAG for managing anticoagulants and antiplatelets. The AI's guidance on the management of cardiovascular medications and antibiotics prophylaxis conformed with ACC-AHA and ASGE guidelines, respectively. Furthermore, in terms of managing diabetes medications, ChatGPT's advice was in accordance with the ADS guidelines. Conclusion: OpenAI's ChatGPT-4 accurately provides answers to pre-colonoscopy questions consistent with medical society guidelines. While AI-powered chatbots like ChatGPT may serve to enhance patient education and autonomy, they should complement, not replace, personalized advice from healthcare professionals.
Ähnliche Arbeiten
Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI
2019 · 8.339 Zit.
Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead
2019 · 8.211 Zit.
High-performance medicine: the convergence of human and artificial intelligence
2018 · 7.614 Zit.
Proceedings of the 19th International Joint Conference on Artificial Intelligence
2005 · 5.776 Zit.
Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI)
2018 · 5.478 Zit.